Why do Immigrants Report Lower Life Satisfaction?

Questions of happiness and well-being have increasingly drawn the attention of health economists. The expectation is that they are an approximate measure of quality of life.

Happiness in surveys is typically reported via a rating scale. For instance, the life satisfaction question at the German Socio-Economic Panel (SOEP): “How satisfied are you at present with your life as a whole?” Respondents answer this question on a scale between 0 and 10, with 0 being the lowest and 10 the highest level of life satisfaction.

It is typically assumed that for a particular latent value of life satisfaction the individual will report the same category (e.g. “8”) regardless of the time of the survey. However, recent research has questioned this, suggesting that a respondent may not use the same evaluation criteria when assessing their life satisfaction at different points in time. For instance, the same level of life satisfaction can be reported as “8” today but as “7” (or lower) in a year from today, if the person becomes more demanding. In this case, we will say that the person will readjust the categories upwards. Similarly, a person who readjusts the categories downwards would for instance score the same level of life satisfaction as “8” today and “9” in a year from today. Therefore, a simplistic analysis between both responses would wrongly conclude that the respondent’s life satisfaction is decreasing or increasing.

A new OHE Research paper by Cubi-Molla and Yaman (2017) explores the potential change in the reported category linked to a given level of life satisfaction. In particular, the authors explore changes in the reporting behaviour of immigrants in Germany 1984-2010.

Previous literature suggests that immigrants’ happiness tends to decrease over time compared to the natives’. Cubi-Molla and Yaman explore the robustness and origin of this finding, and then propose a model to decompose the effect of the number of years since migration into a true change in life satisfaction and a simple change in reporting behaviour. The model exploits the question about how the respondent would, at the point of the interview, rate his or her life satisfaction in the previous year.

The analysis shows that the existence or absence of changes in the reporting behaviour is closely linked to how accurately individuals remember their past life satisfaction. In order to disentangle this relationship, the authors propose a classification of any variable which has an effect on both the latent life satisfaction and on the reporting behaviour into four groups:


Effect on true life satisfaction



Effect on reporting behaviour

Readjusts the categories upwards





Readjusts the categories downwards





As indicated in the table, an adaptive good is a variable that has a positive effect on true life satisfaction, but this effect is not clearly visible in the responses because the individual readjusts the response categories downwards (for example, wealth might be a candidate for an adaptive good: more wealth increases satisfaction, but pushes out the thresholds for classifying oneself as wealthy). Analogous interpretations can be derived for adaptive bad, reinforcing good and reinforcing bad.

The authors found that an excellent recall of past life satisfaction would classify the number of years since migration as a reinforcing good; a not-so-good recall would suggest classifying years since migration as an adaptive bad; assuming a very bad recall would imply years since migration is an adaptive good; under no scenario years since migration would be fit the “reinforcing bad” description.

The model developed in the paper could have many interesting applications for health outcomes. For instance, recent research has questioned the validity of post-test-then-test designs, suggesting that respondents do not interpret in a similar way their quality of life before and after the treatment. In addition, there is no clear evidence that the patient has an accurate recall of pretest functioning when being interviewed after the intervention. The application of the model to this context would permit the estimation of the true quality of life change as a result of a treatment, isolating this effect from any response shift, and controlling for the potential existence of recall bias. The paper encourages more research on this.

Download the full paper here.

For more information please contact Patricia Cubi-Molla at OHE.

Posted in Health Statistics, EQ-5D and PROMs, Research | Tagged Research Papers